The expanding world of enterprise storage
- By Charlotte Adams
- Jul 19, 1998
The enterprise storage market has become an information technology boomtown as customers begin to extend the concept of "enterprise" from the mainframe to other platforms outside the data center and as the amount of data on all platforms continues to grow.
The change is happening as customers expand the role of Unix and Microsoft Corp. Windows NT systems. In the past, these systems tended to run applications for small or midsize workgroups, with only a limited number of people needing access. Now these systems are running more mission-critical applications, so the data needs to be supported by and accessible to the broader operational or enterprise levels.
Additionally, organizations are storing more data at the enterprise level because of the growing use of Internet and intranet applications, data warehouses and data-intensive multimedia information.
"The problem an enterprise faces is that there is so much data," said Jack Scott, a managing partner with Evaluator Group Inc. in Denver. The volume to be stored is growing by 100 percent a year, with the storage industry predicting it will ship around 50 petabytes— that is, 50,000 terabytes— of electronic storage capacity by 2000, Scott said.
Storage vendors are developing new solutions to accommodate the growth. Some vendors have developed technology that allows users to store multiple types of data on a single storage system, while other vendors are developing systems that are designed to give users fast access to data stored across multiple platforms on a network.
The two approaches might be described as "storage-centric" and "network-attached storage."
With a storage-centric approach, an agency crams multiple data types into a single, extremely large system. This segment is dominated by traditional mainframe-class storage vendors.
EMC Corp.'s newest Symmetrix 5700/3700 system, for example, comes with an array of 47G drives for up to 6 tera-
bytes capacity, and it supports mainframe, multiple Unix flavors and Windows NT storage in the same enclosure. The company is a major player in the Defense Logistics Agency and the Treasury, Justice and Commerce departments, said Bruce Triner, vice president for EMC's federal operations in Bethesda, Md.
Hitachi Data Systems Inc.'s Freedom 7700E also can handle multiple data types. It features 3-terabyte capacity, 10G upfront cache, for better processing performance and the ability to handle 30,000 input/output operations. Hitachi customers include Defense Department megacenters, the Patent and Trademark Office and the Social Security Administration, according to Tom Green, supervisor of technical support at Vion Corp., Hitachi's federal reseller.
Although vendors such as EMC, Hitachi and IBM Corp. support heterogeneous storage, such requirements rarely surface in federal solicitations, said Sherry Straley, high-end storage brand manager at IBM Global Government Industry in Bethesda, Md.
Network-attached storage, pushed by Sun Microsystems Inc., Data General Corp. and other open-systems vendors, takes a different tack. This approach builds on the Fibre Channel connectivity standard, which uses optical fiber to provide up to 100 megabytes/sec data transfer rates. Vendors create a private, Fibre Channel-based "storage-area network," for high data transfers between servers and storage, while also offloading storage traffic from the regular local-area network.
Because of the high-speed links, storage can be distributed across an organization without slowing down performance, according to vendors.
Also, a single storage system does not have to be everything to everybody. The more distributed approach has the advantages of easier growth and almost infinite scalability, and it gives users flexibility to put servers and storage systems where they prefer, Scott said.
This architecture, which EMC and Hitachi also support, is just beginning to enter the market.
Sun uses Fibre Channel to support "desktop-to-enterprise" storage, including both a high-end, heterogeneous storage system, the A7000, and numerous Unix boxes, such as the all-Fibre Channel A5000, said Jeff Allen, senior director of marketing for network storage with Sun in Newark, Calif.
The A7000 offers up to 32 host connections, supporting 2.9 terabytes of data with up to 4G cache. According to some analysts, it comes closest to the data-sharing grail, whereby different environments share a single data view.
A major A7000 customer is the Army Personnel Command, which needs to maintain up-to-date information on the more than 300,000 enlisted and 100,000 officers on active duty, said John Leahy, group manager for government affairs with Sun Federal.
A smaller-scale operation at a naval aviation test facility plans to employ a cabinet with two 120G A5000 arrays, with room for twice that amount. The A5000 offers Fibre Channel connectivity to the application's Enterprise 6000 servers, said a Navy systems administrator. The "miniature data center" works on a relatively small scale, however, with a maximum of 20 simultaneous users and generating a flow of "a few gigabytes" of data a week, the administrator said.
The advantage of a Fibre Channel-based, network-attached system is that it enables users to "to scale storage systems to applications' requirements rather than having to buy it in large chunks," said Joel Reich, group product line manager with Data General's Clariion Storage Division, Southboro, Mass.
Data General allows "dynamic upgrades," meaning that customers can add capacity to an existing system without turning it off, he said. The company sells storage through Storage Technology Corp., Hewlett-Packard Co. and Silicon Graphics Inc. Data General's FC5000 system is used in numerous federal satellite imagery applications.
The network-attached storage approach is catching on, but at the moment is "better for smaller problems," said David Vellante, a senior vice president at IDC, Framingham, Mass. However, the idea of "[using] Fibre Channel to build storage-area networks that are as functional as an EMC box is vaporware," he said.
However, the increasing complexity of the enterprise storage environment brings with it more complex data management problems, industry observers said.
"Data centers are getting a hodgepodge of Unix and [Windows] NT storage," as well as mainframe boxes, said John Young, vice president of enterprise systems planning with the Clipper Group, Wellesley, Mass.
The trouble with open-systems storage is that the "tools available are not as robust as in the traditional [Multiple Virtual Storage mainframe] environment," said Bill Finefield, chief information officer for the Navy Exchange Services Command, Virginia Beach, Va. "Today Unix looks like an MVS shop 20 years ago— it's very labor-intensive."
Nexcom has consolidated seven data centers into one. At the moment Finefield has 1.5 terabytes of Unix data stored on two EMC boxes. The command needs bandwidth as well as storage space, Finefield said, because the data center crunches batch loads 24 hours a day, every day.
Backup is also a sore point, said Bob Cosby, the direct access storage device (DASD) administrator at the Agriculture Department's National Financial Center in New Orleans, which has 4.2 terabytes of DASD online and handles payroll check processing for multiple agencies.
Each storage system costs $300,000 to $1 million, yet there is no cross-platform, "quick snapshot" backup capability, he said. The data center uses Innovation Data Processing software to back up certain disks across platforms, but the capability is not instantaneous. Cosby has avoided vendor lock-in by purchasing mainframe storage from IBM, Hitachi and EMC, as well as an EMC box for Unix.
Cosby is also "very skeptical," about how well MVS data and Unix data co-exist on a single subsystem. "We're a very conservative shop. We've been on the bleeding edge before and don't want to go there again," he said. Cost considerations related to integrating Unix applications and data into the MVS environment also dictate data segregation.
Still, management software is progressing. EMC, for example, offers TimeFinder, which can create a second copy of data and "split it off from the first" within the same Symmetrix box for data backup or for use in data warehouses, said Doug Fierro, manager of product marketing at EMC. Also, the company has introduced a product called Power Path to help Unix customers maintain high bandwidth by balancing a data workload across up to 32 SCSI connections between servers and EMC storage.
IBM also stresses software as part of its emerging Seascape storage blueprint. StorWatch will be the "management software glue" for the architecture, including tools to view, manage and control resources, said Scott Kozar, program manager for disk marketing with IBM's Storage Division, White Plains, N.Y. The blueprint calls for a slew of tools to address configuration, capacity, maintenance, performance and asset management as well other areas.
Adams is a free-lance writer based in Alexandria, Va. She can be reached at [email protected]